From uchinews!linac!uwm.edu!wupost!howland.reston.ans.net!agate!netsys!pagesat!olivea!charnel!psgrain!ee.und.ac.za!csir.co.za!nuustak!acooper Wed Mar 3 19:46:51 CST 1993 Article: 35405 of alt.folklore.computers Path: uchinews!linac!uwm.edu!wupost!howland.reston.ans.net!agate!netsys!pagesat!olivea!charnel!psgrain!ee.und.ac.za!csir.co.za!nuustak!acooper From: acooper@nuustak.csir.co.za (Antony Cooper) Newsgroups: alt.folklore.computers Subject: Re: Eliza and Racter... PD? Message-ID: Date: 3 Mar 93 18:37:28 GMT Organization: CSIR, South AFrica Lines: 43 NNTP-Posting-Host: nuustak.csir.co.za X-Disclaimer: The opinions expressed below are purely personal X-Disclaimer: and in no way represent the official view of CSIR. Status: R Thus spake wbdst+@pitt.edu (William B Dwinnell): >There are a myriad such programs [as Eliza and RACTER]. At a recent >event (within the last year or 2), at the Boston Computer Museum >(if you haven't been there, you should check it out), there were >several such conversational programs, all competing with humans in a >modified Turing test (yes, that forever vaunted test), and interestingly, 1 human >passed herself off as a computer (I think it was a 'her'), she was an >English expert I believe (hmmm... :) ), and one computer program which >was able to successfully fool the human judges! The First Annual Loebner Prize Competition, to which you are referring, is described in detail in the article "Can machines think?", on pages 80-95 in the Summer 1992 issue of AI Magazine, Vol 13, No 2. The article is very interesting, and includes the actual transcript of the session that won the competition. The competition was held on 8 November 1991. The Loebner Prize Competition is a scaled-down version of the actual Turing Test. The article includes a description of why the competition was structured as it was, and I recommend that you look for it. The computers performed relatively poorly for the most part, which is what they expected. The winning entry, "Whimsical conversation" by Joseph Weintraub of New York, fooled half of the judges into thinking that it was human, and two of the ten judges rated it as more human than one or both of the competing humans. The two main features of "Whimsical conversation" were that it simulated typing errors well, and that Weintraub selected his topic well (each of the competitors discussed only a narrow topic, one of the restrictions on the proper Turing Test), namely that of a jester - one can get away with quite a bit when joking away, because people are more tolerant of such a "person". Cynthia Clay was mistaken for a computer by three of the judges. Her topic was Shakespeare's plays, on which she is an expert, and she was able to quote length passages verbatim, which made several judges say that her replies were too expert to be from a human! The second Loebner Prize Competition was held on 17 November 1992. Has anyone heard anything about how well the contestants fared there? Antony -- Antony Cooper | Voice: +27 12 841 4121 acooper@nuustak.csir.co.za | Fax: +27 12 841 3037 INFOTEK, CSIR, Box 395, Pretoria, 0001, South Africa | ICBM: 25 45S 28 16E